Transformer substation environment risk monitoring and early warning method and device based on multi-dimensional data fusion, electronic equipment and storage medium

Abstract

The invention discloses a transformer substation environment risk monitoring and early warning method and device based on multi-dimensional data fusion, electronic equipment and a storage medium, and belongs to the field of transformer substation environment monitoring and early warning, and the method comprises the steps: obtaining the noise, environment weather, electromagnetic radiation and monitoring video data of a transformer substation; calculating a noise prediction value of the target residential area based on an acoustic propagation attenuation model, and comparing the noise prediction value with a noise limit value to generate a noise risk; calculating an electromagnetic radiation prediction value based on the electromagnetic radiation data and the distance parameter, and comparing the electromagnetic radiation prediction value with a control limit value to generate an electromagnetic radiation risk; the risk of waste generation due to abnormal waste discharge is identified through video analysis; and taking the maximum value in the three risks as a comprehensive risk value, comparing the comprehensive risk value with a risk threshold to determine an environmental risk level, and generating an alarm control signal and an operation and maintenance strategy. By implementing the method, the problem that the influence value of the target residential area is difficult to accurately predict due to the fact that the environmental meteorological parameters and the propagation distance parameters are not incorporated into a calculation model in the prior art can be solved.

Description

Technical Field

[0001] This invention relates to the field of substation environmental monitoring and early warning technology, specifically to a substation environmental risk monitoring and early warning method, device, electronic equipment, and storage medium based on multi-dimensional data fusion. Background Technology

[0002] Building an efficient and accurate environmental risk monitoring and early warning system for substations is crucial for mitigating the "NIMBY effect," protecting residents' health, and ensuring the safe operation of the power grid.

[0003] However, existing substation environmental monitoring technologies still have significant shortcomings in terms of accuracy and intelligent sensing. The main problems are: current acoustic and electromagnetic environment monitoring is mostly limited to simple data collection at substation boundaries, lacking a comprehensive calculation mechanism that incorporates environmental meteorological parameters and propagation distance parameters into the acoustic propagation attenuation model. This makes it difficult to accurately calculate the actual impact values ​​at target residential areas from boundary monitoring data, resulting in discrepancies between early warning results and residents' actual experiences. Simultaneously, existing video surveillance focuses primarily on image recording, lacking the ability to automatically quantify and identify waste emissions based on image feature analysis, making it difficult to meet the real-time intelligent sensing needs of substation environmental risks. Summary of the Invention

[0004] This invention provides a method, device, electronic device, and storage medium for monitoring and early warning of environmental risks in substations based on multi-dimensional data fusion. It can solve the problems in the prior art, such as the difficulty in accurately predicting the actual impact value of target residential areas due to the failure to incorporate environmental meteorological parameters and propagation distance parameters into the calculation model, and the lack of automated identification capability for waste emissions.

[0005] One embodiment of the present invention provides a method for monitoring and early warning of environmental risks in substations based on multi-dimensional data fusion, comprising: Acquire noise data, environmental meteorological parameter data, electromagnetic radiation data, and monitoring video data of the substation; Based on the noise data, the environmental meteorological parameter data, and the preset propagation distance parameter, the predicted noise value at the target residential area is calculated using an acoustic propagation attenuation model; the predicted noise value is compared with the preset noise emission limit to generate a noise risk value; wherein, the propagation distance parameter represents the distance between the substation and the target residential area; Based on the electromagnetic radiation data, the propagation distance parameter, and the preset monitoring terminal distance parameter, the predicted electromagnetic radiation value at the target residential area is calculated; the predicted electromagnetic radiation value is compared with the preset electromagnetic control limit to generate an electromagnetic radiation risk value; wherein, the monitoring terminal distance parameter represents the distance between the electromagnetic radiation data acquisition location and the radiation source in the substation. Image feature analysis is performed on the monitoring video data to identify abnormal waste discharge situations and generate waste risk values; The maximum value among the noise risk value, the electromagnetic radiation risk value, and the waste risk value is selected and determined as the comprehensive risk value; the comprehensive risk value is compared with a preset risk classification threshold to determine the substation environmental risk level; and corresponding alarm control signals and operation and maintenance strategies are generated based on the substation environmental risk level.

[0006] Furthermore, based on the noise data, the environmental meteorological parameter data, and the preset propagation distance parameter, the predicted noise value at the target residential area is calculated using an acoustic propagation attenuation model; the predicted noise value is compared with the preset noise emission limit to generate a noise risk value, including: Based on the propagation distance parameter, the geometric divergence attenuation of the sound wave during propagation is calculated; The atmospheric absorption attenuation coefficient is determined based on the environmental meteorological parameter data, and the atmospheric absorption attenuation of the sound wave during propagation is calculated using the propagation distance parameter and the atmospheric absorption attenuation coefficient. The noise data is attenuated and corrected using the geometric divergence attenuation and the atmospheric absorption attenuation to obtain the predicted noise value at the target residential area. If the predicted noise value is greater than the preset noise emission limit, then the first difference between the predicted noise value and the noise emission limit is calculated, and the ratio of the first difference to the noise emission limit is determined as the noise risk value. If the predicted noise value is less than or equal to the noise emission limit, then the noise risk value is set to zero.

[0007] Furthermore, based on the electromagnetic radiation data, the propagation distance parameter, and the preset monitoring end distance parameter, the predicted electromagnetic radiation value at the target residential area is calculated; the predicted electromagnetic radiation value is compared with the preset electromagnetic control limit to generate an electromagnetic radiation risk value, including: Based on the preset monitoring end distance parameter and the propagation distance parameter, an electromagnetic attenuation coefficient characterizing the distance attenuation characteristics is constructed; The electromagnetic radiation data is attenuated and corrected using the electromagnetic attenuation coefficient to obtain the predicted electromagnetic radiation value at the target residential area. If the predicted electromagnetic radiation value is greater than the preset electromagnetic control limit, then the second difference between the predicted electromagnetic radiation value and the electromagnetic control limit is calculated, and the ratio of the second difference to the electromagnetic control limit is determined as the electromagnetic radiation risk value. If the predicted electromagnetic radiation value is less than or equal to the electromagnetic control limit, then the electromagnetic radiation risk value is set to zero.

[0008] Furthermore, the step of performing image feature analysis on the monitoring video data to identify abnormal waste discharge situations and generate waste risk values ​​includes: The frame image to be detected is extracted from the monitoring video data according to a preset sampling frequency; For each frame image to be detected, a background reference image that matches the lighting environment features of the current frame image to be detected is obtained, and the current frame image to be detected and the background reference image are subjected to a difference operation to extract the foreground difference region; Calculate the color mean and texture complexity of the foreground difference region; If the average color value of the foreground difference region falls within the preset waste color range, and the texture complexity of the foreground difference region is less than the preset liquid texture threshold, then it is determined that there is a suspected waste region in the current frame image to be detected. The system counts the number of consecutive frames of images identified as having suspected waste areas. If the number of consecutive frames of images with suspected waste areas reaches a preset alarm threshold, the frame of images that exactly reaches the alarm threshold is selected as the target image. The area ratio of the suspected waste area in the target image is calculated, and the area ratio is determined as the waste risk value. Otherwise, assign a waste risk value of zero.

[0009] Furthermore, the step of performing image feature analysis on the monitoring video data to identify abnormal waste discharge situations and generate waste risk values ​​includes: The serialized frame images to be detected are extracted from the monitoring video data according to a preset sampling frequency; Each of the images to be detected is sequentially input into a preset waste identification neural network model, so that the waste identification neural network model generates the recognition confidence and waste target area of ​​each image to be detected based on the images to be detected. A counter is used to count the number of frames of the image to be detected whose recognition confidence is continuously greater than a preset effective threshold; wherein, when the recognition confidence of the current frame image is greater than the effective threshold, the counter is incremented by one; when the recognition confidence of the current frame image is less than or equal to the effective threshold, the counter is reset to zero. When the counter value accumulates to a preset alarm threshold, an abnormal waste discharge is identified, and the corresponding frame image to be detected that just causes the counter value to reach the alarm threshold is selected as the target image. Calculate the area ratio of the waste target region in the target image, and determine the area ratio as the waste risk value; If the counter does not reach the alarm threshold, the waste risk value will be set to zero.

[0010] Furthermore, the training of the waste identification neural network model includes: Obtain a waste sample training dataset; wherein, the training dataset includes several substation scene sample images containing various types of waste, and each substation scene sample image is labeled with the real area coordinates and real category label of the waste; The waste sample training dataset is divided into several batches of training samples according to a preset batch size; Each batch of training samples is input into the waste identification neural network model to be trained for iterative training until the preset number of training rounds is met; wherein, in each iteration, the waste identification neural network model to be trained outputs the predicted waste area coordinates and the predicted category confidence based on the current batch of training samples. Based on the predicted waste area coordinates and the corresponding real area coordinates, a location loss value is calculated using a location regression loss function; based on the predicted category confidence and the corresponding real category label, a classification loss value is calculated using a classification loss function; the location loss value and the classification loss value are weighted and summed to obtain a comprehensive loss value, and a preset optimizer is used to update the network parameters of the waste identification neural network model based on the comprehensive loss value.

[0011] Furthermore, the risk classification threshold includes a first risk threshold, a second risk threshold, and a third risk threshold; the first risk threshold is less than the second risk threshold, and the second risk threshold is less than the third risk threshold; The step of comparing the comprehensive risk value with a preset risk classification threshold to determine the environmental risk level of the substation includes: If the comprehensive risk value is less than the first risk threshold, the environmental risk level of the substation is determined to be acceptable. If the comprehensive risk value is greater than or equal to the first risk threshold and less than the second risk threshold, then the environmental risk level of the substation is determined to be a minor risk level. If the comprehensive risk value is greater than or equal to the second risk threshold and less than the third risk threshold, then the environmental risk level of the substation is determined to be a medium risk level. If the comprehensive risk value is greater than or equal to the third risk threshold, the environmental risk level of the substation is determined to be a severe risk level.

[0012] Based on the above method embodiments, the present invention provides corresponding apparatus embodiments.

[0013] An embodiment of the present invention provides a substation environmental risk monitoring and early warning device based on multi-dimensional data fusion, comprising: a data acquisition module, a noise risk assessment module, an electromagnetic radiation risk assessment module, a waste risk assessment module, and a comprehensive early warning and control module; The data acquisition module is used to acquire noise data, environmental meteorological parameter data, electromagnetic radiation data, and monitoring video data of the substation. The noise risk assessment module is used to calculate the predicted noise value at the target residential area based on the noise data, the environmental meteorological parameter data, and the preset propagation distance parameter, using an acoustic propagation attenuation model; compare the predicted noise value with the preset noise emission limit to generate a noise risk value; wherein, the propagation distance parameter represents the distance between the substation and the target residential area; The electromagnetic radiation risk assessment module is used to calculate the predicted electromagnetic radiation value at the target residential area based on the electromagnetic radiation data, the propagation distance parameter, and the preset monitoring end distance parameter; compare the predicted electromagnetic radiation value with the preset electromagnetic control limit to generate an electromagnetic radiation risk value; wherein, the monitoring end distance parameter represents the distance between the electromagnetic radiation data acquisition location and the radiation source in the substation. The waste risk assessment module is used to perform image feature analysis on the monitoring video data, identify abnormal waste discharge situations, and generate waste risk values. The integrated early warning and control module is used to select the maximum value among the noise risk value, the electromagnetic radiation risk value, and the waste risk value, and determine the maximum value as the integrated risk value; compare the integrated risk value with a preset risk classification threshold to determine the substation environmental risk level; and generate corresponding alarm control signals and operation and maintenance strategies based on the substation environmental risk level.

[0014] Based on the above method embodiments, the present invention provides corresponding electronic device embodiments.

[0015] An embodiment of the present invention provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the substation environmental risk monitoring and early warning method based on multi-dimensional data fusion as described in any of the above-described method embodiments.

[0016] Based on the above method embodiments, the present invention provides corresponding storage medium embodiments.

[0017] One embodiment of the present invention provides a storage medium storing a computer program thereon, wherein, when the computer program is running, it controls the device where the storage medium is located to execute any of the above-described method embodiments of the substation environmental risk monitoring and early warning method based on multi-dimensional data fusion.

[0018] Compared with the prior art, the present invention has the following beneficial effects: This invention provides a method, device, electronic device, and storage medium for substation environmental risk monitoring and early warning based on multi-dimensional data fusion. The method acquires noise data, environmental meteorological parameter data, electromagnetic radiation data, and monitoring video data from the substation; based on the noise data, environmental meteorological parameters, and propagation distance parameters, it calculates the predicted noise value of the target residential area using an acoustic propagation attenuation model and compares it with noise emission limits to generate a noise risk value; based on the electromagnetic radiation data, propagation distance parameters, and monitoring end distance parameters, it calculates the predicted electromagnetic radiation value of the target residential area and compares it with electromagnetic control limits to generate an electromagnetic radiation risk value; it performs image feature analysis on the monitoring video data to identify abnormal waste emissions and generate a waste risk value; it selects the maximum value among the three risk values ​​(noise, electromagnetic radiation, and waste) as a comprehensive risk value and compares it with a preset risk classification threshold to determine the substation's environmental risk level; and it generates alarm control signals and corresponding operation and maintenance strategies based on the risk level.

[0019] This invention calculates predicted noise levels at target residential areas using an acoustic propagation attenuation model based on environmental meteorological parameters and propagation distance parameters. It also calculates predicted electromagnetic radiation levels by combining these parameters with distance parameters. This achieves accurate estimation of the actual impact on target residential areas from substation monitoring data, effectively solving the problem of discrepancies between warning results and residents' actual experiences caused by existing technologies that do not incorporate meteorological factors and spatial distance into their calculation models. Simultaneously, this invention utilizes image feature analysis technology to achieve automated quantitative identification of waste emissions, generating risk values ​​and alarm strategies accordingly. This effectively addresses the problem of existing technologies lacking automated waste emission identification capabilities, thus failing to meet real-time intelligent sensing requirements. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating a substation environmental risk monitoring and early warning method based on multi-dimensional data fusion, provided by an embodiment of the present invention.

[0021] Figure 2 This is a schematic diagram of the structure of a substation environmental risk monitoring and early warning device based on multi-dimensional data fusion, provided in an embodiment of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] like Figure 1 As shown, to address the problems in existing technologies where the calculation model does not incorporate environmental meteorological parameters and propagation distance parameters, making it difficult to accurately predict the actual impact on target residential areas, and the lack of automated identification capabilities for waste emissions, an embodiment of the present invention provides a substation environmental risk monitoring and early warning method based on multi-dimensional data fusion, comprising at least the following steps: Step S1: Obtain noise data, environmental meteorological parameter data, electromagnetic radiation data, and monitoring video data of the substation; Specifically, in the implementation of the substation environmental risk monitoring and early warning method based on multi-dimensional data fusion, the first step is data acquisition. Various environmental data are collected in real time by multi-dimensional monitoring equipment deployed at the substation site and transmitted to the data processing terminal. The multi-dimensional monitoring equipment specifically includes a noise monitor, environmental meteorological parameter sensors, an electromagnetic environment monitor, and a surveillance video camera. The noise monitor is used to monitor the noise level of the substation and its surroundings in real time and converts the collected acoustic signals into digital signals. Here, the noise data measured by the noise monitor is defined as the boundary noise sound pressure level. The unit is decibels (dB), and the boundary noise sound pressure level is... This will be used as the sound source intensity input in the subsequent acoustic propagation model. The environmental meteorological parameter sensor is used to collect real-time meteorological conditions in the substation area. The environmental meteorological parameter data includes at least temperature data, humidity data, air pressure data, and wind speed data. The temperature and humidity data are used to determine the atmospheric absorption attenuation coefficient during sound wave propagation, and the wind speed data is used to correct for the influence of the sound wave propagation direction. The electromagnetic environment monitoring instrument is used to monitor the electromagnetic radiation intensity of the substation and its surroundings, and uses the collected electromagnetic field data as the electromagnetic radiation data. Here, the electromagnetic radiation data measured by the electromagnetic environment monitoring instrument is defined as the boundary electromagnetic radiation quantity. The unit is milligauss (mG), and the boundary electromagnetic radiation quantity This will serve as the basis for subsequent electromagnetic attenuation calculations. The monitoring video camera is used to continuously capture images of the substation and its surrounding environment, obtaining high-resolution real-time video stream data or periodically captured frame image data, which will be used as the monitoring video data for subsequent image feature analysis and waste identification.

[0024] In one embodiment, to ensure comprehensive coverage and accurate capture of environmental risks at the substation, the noise monitor, the environmental meteorological parameter sensor, the electromagnetic environment monitor, and the surveillance video camera are integrated into a single monitoring terminal. Four such monitoring terminals are configured along the boundary of the substation, installed at the easternmost, westernmost, southernmost, and northernmost points of the substation boundary, respectively. Each terminal is securely installed using cement piles or other rigid fixing structures to ensure the stability and consistency of data acquisition locations. The monitoring terminal synchronously transmits the collected noise data, environmental meteorological parameter data, electromagnetic radiation data, and surveillance video data to a backend processor for subsequent calculation and analysis via wired or wireless networks. By acquiring multi-dimensional environmental data of the substation in one go through the above steps, a foundation for environmental perception encompassing sound, light, electricity, and air can be built. This ensures sufficient and necessary real-time data support for subsequent target point prediction using physical attenuation models and waste identification using artificial intelligence algorithms, thereby effectively improving the timeliness and accuracy of environmental risk monitoring.

[0025] Step S2: Based on the noise data, the environmental meteorological parameter data, and the preset propagation distance parameter, calculate the predicted noise value at the target residential area using an acoustic propagation attenuation model; compare the predicted noise value with the preset noise emission limit to generate a noise risk value; wherein, the propagation distance parameter represents the distance between the substation and the target residential area; In a preferred embodiment, the step of calculating the predicted noise value at the target residential area using an acoustic propagation attenuation model based on the noise data, the environmental meteorological parameter data, and a preset propagation distance parameter; and comparing the predicted noise value with a preset noise emission limit to generate a noise risk value, includes: Based on the propagation distance parameter, the geometric divergence attenuation of the sound wave during propagation is calculated; The atmospheric absorption attenuation coefficient is determined based on the environmental meteorological parameter data, and the atmospheric absorption attenuation of the sound wave during propagation is calculated using the propagation distance parameter and the atmospheric absorption attenuation coefficient. The noise data is attenuated and corrected using the geometric divergence attenuation and the atmospheric absorption attenuation to obtain the predicted noise value at the target residential area. If the predicted noise value is greater than the preset noise emission limit, then the first difference between the predicted noise value and the noise emission limit is calculated, and the ratio of the first difference to the noise emission limit is determined as the noise risk value. If the predicted noise value is less than or equal to the noise emission limit, then the noise risk value is set to zero.

[0026] Specifically, in the implementation of the substation environmental risk monitoring and early warning method based on multi-dimensional data fusion, after acquiring the noise data and the environmental meteorological parameter data, the process enters the noise prediction and risk assessment stage. The method utilizes a preset acoustic propagation attenuation model, combined with the noise data, the environmental meteorological parameter data, and a preset propagation distance parameter, to calculate the predicted noise value at the target residential area. The propagation distance parameter (denoted as...) The propagation distance parameter represents the spatial transmission distance between the sound source (i.e., the monitoring point location) of the substation and the target residential area. These are static constants that are pre-determined and stored in the system. The acoustic propagation attenuation model comprehensively considers the geometric divergence attenuation and atmospheric absorption attenuation during sound wave propagation to achieve accurate extrapolation of the noise level at a distance.

[0027] In a preferred embodiment, the specific process of calculating the predicted noise value at the target residential area is as follows: First, based on the propagation distance parameter... Calculate the geometrical divergence attenuation of the sound wave during propagation (denoted as ). Since sound waves, as spherical waves, attenuate with increasing distance when propagating in a free sound field, the geometrical divergence attenuation... With the propagation distance parameter The value is proportional to the logarithm of the value. Secondly, the atmospheric absorption attenuation coefficient is determined based on the aforementioned environmental meteorological parameter data. Specifically, current air temperature and relative humidity data are extracted from the environmental meteorological parameter data, and the atmospheric absorption attenuation coefficient (denoted as ν) at the corresponding frequency is obtained by querying or calculating according to acoustic standards (such as ISO 9613-1). The atmospheric absorption attenuation coefficient This characterizes the sound energy loss due to air absorption per unit distance. Next, the propagation distance parameter is used... and the atmospheric absorption attenuation coefficient Calculate the atmospheric absorption attenuation of sound waves during propagation (denoted as ). The calculation formula can be expressed as: In the formula, This represents atmospheric absorption attenuation, expressed in decibels (dB). The atmospheric absorption attenuation coefficient is expressed in decibels per kilometer (dB / km). This is the propagation distance parameter, with the unit being meters (m).

[0028] After calculating the various attenuation values, the geometrical divergence attenuation value is used. and the atmospheric absorption attenuation amount For the noise data (i.e., boundary noise sound pressure level) The attenuation correction is performed to obtain the predicted noise value at the target residential area (denoted as ). The specific prediction calculation formula is as follows: in, The predicted noise level at the target residential area is expressed in decibels (dB). Noise data measured at the substation boundary, in decibels (dB). This is the geometrical divergence attenuation, measured in decibels (dB). This is achieved by introducing the atmospheric absorption attenuation. By modifying the traditional distance attenuation model, errors caused by changes in meteorological conditions in noise propagation prediction can be eliminated.

[0029] Next, the noise risk value generation step is performed. The calculated predicted noise value is then used. Compared with the preset noise emission limit (denoted as The noise emission limits are compared. This is usually set according to the limits for various acoustic environmental functional zones specified in national acoustic environmental quality standards (such as GB3096-2008). If the comparison results show the predicted noise value... Greater than the noise emission limit If the noise level exceeds the limit, it is determined that there is a risk of noise exceeding the standard, and the degree of risk needs to be further calculated and quantified. At this point, the predicted noise value is calculated. With the aforementioned noise emission limits The first difference, and compare the first difference with the noise emission limit. The ratio of is determined as the noise risk value (denoted as ). The noise risk value The calculation formula is as follows: in, This is a dimensionless noise risk value used to characterize the relative degree of noise exceeding the standard. If the comparison results show the predicted noise value... Less than or equal to the noise emission limit If the ambient noise level is within a safe range, the noise risk value is directly set. The value is assigned to zero. Through the above steps, real-time monitoring data of the substation boundary can be transformed into quantitative risk indicators for surrounding sensitive target areas, realizing dynamic and accurate assessment of substation noise environmental risks and avoiding false alarms or omissions caused by relying solely on boundary data.

[0030] Step S3: Calculate the predicted electromagnetic radiation value at the target residential area based on the electromagnetic radiation data, the propagation distance parameter, and the preset monitoring terminal distance parameter; compare the predicted electromagnetic radiation value with the preset electromagnetic control limit to generate an electromagnetic radiation risk value; wherein, the monitoring terminal distance parameter represents the distance between the electromagnetic radiation data acquisition location and the radiation source within the substation. In a preferred embodiment, the step of calculating the predicted electromagnetic radiation value at the target residential area based on the electromagnetic radiation data, the propagation distance parameter, and a preset monitoring end distance parameter; and comparing the predicted electromagnetic radiation value with a preset electromagnetic control limit to generate an electromagnetic radiation risk value, includes: Based on the preset monitoring end distance parameter and the propagation distance parameter, an electromagnetic attenuation coefficient characterizing the distance attenuation characteristics is constructed; The electromagnetic radiation data is attenuated and corrected using the electromagnetic attenuation coefficient to obtain the predicted electromagnetic radiation value at the target residential area. If the predicted electromagnetic radiation value is greater than the preset electromagnetic control limit, then the second difference between the predicted electromagnetic radiation value and the electromagnetic control limit is calculated, and the ratio of the second difference to the electromagnetic control limit is determined as the electromagnetic radiation risk value. If the predicted electromagnetic radiation value is less than or equal to the electromagnetic control limit, then the electromagnetic radiation risk value is set to zero.

[0031] Specifically, in the implementation process of the substation environmental risk monitoring and early warning method, after completing the noise risk assessment, or in parallel with the noise assessment, the steps of electromagnetic radiation prediction and risk quantification are performed. The method derives the predicted electromagnetic radiation value at the target residential area through a physical calculation model based on the acquired electromagnetic radiation data, the preset propagation distance parameter, and the preset monitoring end distance parameter. The electromagnetic radiation data refers to the electromagnetic field strength value collected in real time by the monitoring terminal at the substation boundary, in milligauss (mG); the propagation distance parameter represents the straight-line distance between the center of the radiation source within the substation and the sensitive point in the target residential area; the monitoring end distance parameter represents the straight-line distance between the location where the electromagnetic radiation data is collected (i.e., the location of the cement pile where the monitoring terminal is installed) and the center of the radiation source within the substation. Since the electromagnetic radiation source is usually located inside the substation (such as a transformer or high-voltage coil), while the monitoring point is located at the boundary, the monitoring end distance parameter is usually smaller than the propagation distance parameter.

[0032] In a preferred embodiment, the specific process of calculating the predicted electromagnetic radiation value at the target residential area follows the inverse square law of electromagnetic wave propagation. First, based on the preset monitoring end distance parameter and the propagation distance parameter, an electromagnetic attenuation coefficient characterizing the distance attenuation properties (denoted as...) is constructed. Specifically, the ratio of the monitoring end distance parameter to the propagation distance parameter is calculated, and the square of the ratio is determined as the electromagnetic attenuation coefficient. The electromagnetic attenuation coefficient The calculation formula is as follows: in, The electromagnetic attenuation coefficient is a dimensionless coefficient. For the distance parameter of the monitoring end, This is the propagation distance parameter. Subsequently, the electromagnetic attenuation coefficient is used... Regarding the electromagnetic radiation data Attenuation correction is performed by extrapolating the measured values ​​at the boundary to the distant receptor location through multiplication operations, thereby obtaining the predicted electromagnetic radiation value at the target residential area (denoted as ). The calculation formula is as follows: in, The value represents the predicted electromagnetic radiation at the target residential area, expressed in milligauss (mG). This calculation process is based on the physical property that point source radiation energy decays with the square of the distance, enabling the scientific inversion of far-field environmental conditions using near-field monitoring data.

[0033] After obtaining the predicted electromagnetic radiation value, the electromagnetic radiation risk value generation step is performed. The predicted electromagnetic radiation value is then compared with a preset electromagnetic control limit (denoted as...). The electromagnetic control limits are compared. The electromagnetic radiation value is set according to the national standard "Electromagnetic Environment Control Limits" (GB8702-2014) or other relevant power industry standards (e.g., set at 4000 mG or public exposure control limits). If the comparison results show the predicted electromagnetic radiation value... Greater than the electromagnetic control limit If this is the case, then it is determined that the current electromagnetic environment poses a risk of exceeding the standard, and the specific degree of risk needs to be calculated. At this point, the predicted electromagnetic radiation value is calculated. With the electromagnetic control limit The second difference, and compare the second difference with the electromagnetic control limit. The ratio of these values ​​is determined as the electromagnetic radiation risk value (denoted as ). The electromagnetic radiation risk value. The calculation formula is as follows: in, This is a dimensionless numerical value used to quantify the degree to which electromagnetic radiation exceeds the standard. If the comparison results show the predicted electromagnetic radiation value... Less than or equal to the electromagnetic control limit value This indicates that the current electromagnetic environment is within a safe range, and the electromagnetic radiation risk value can be directly set to... The value is assigned to zero. Through the above steps, based on fixed boundary monitoring points, the electromagnetic environment quality of sensitive targets at different distances can be dynamically assessed, providing a quantitative basis for the compliance management of the electromagnetic environment.

[0034] Step S4: Perform image feature analysis on the monitoring video data to identify abnormal waste discharge situations and generate waste risk values; In a preferred embodiment, the step of performing image feature analysis on the monitoring video data to identify abnormal waste discharge situations and generate waste risk values ​​includes: The frame image to be detected is extracted from the monitoring video data according to a preset sampling frequency; For each frame image to be detected, a background reference image that matches the lighting environment features of the current frame image to be detected is obtained, and the current frame image to be detected and the background reference image are subjected to a difference operation to extract the foreground difference region; Calculate the color mean and texture complexity of the foreground difference region; If the average color value of the foreground difference region falls within the preset waste color range, and the texture complexity of the foreground difference region is less than the preset liquid texture threshold, then it is determined that there is a suspected waste region in the current frame image to be detected. The system counts the number of consecutive frames of images identified as having suspected waste areas. If the number of consecutive frames of images with suspected waste areas reaches a preset alarm threshold, the frame of images that exactly reaches the alarm threshold is selected as the target image. The area ratio of the suspected waste area in the target image is calculated, and the area ratio is determined as the waste risk value. Otherwise, assign a waste risk value of zero.

[0035] In a preferred embodiment, the step of performing image feature analysis on the monitoring video data to identify abnormal waste discharge situations and generate waste risk values ​​includes: The serialized frame images to be detected are extracted from the monitoring video data according to a preset sampling frequency; Each of the images to be detected is sequentially input into a preset waste identification neural network model, so that the waste identification neural network model generates the recognition confidence and waste target area of ​​each image to be detected based on the images to be detected. A counter is used to count the number of frames of the image to be detected whose recognition confidence is continuously greater than a preset effective threshold; wherein, when the recognition confidence of the current frame image is greater than the effective threshold, the counter is incremented by one; when the recognition confidence of the current frame image is less than or equal to the effective threshold, the counter is reset to zero. When the counter value accumulates to a preset alarm threshold, an abnormal waste discharge is identified, and the corresponding frame image to be detected that just causes the counter value to reach the alarm threshold is selected as the target image. Calculate the area ratio of the waste target region in the target image, and determine the area ratio as the waste risk value; If the counter does not reach the alarm threshold, the waste risk value will be set to zero.

[0036] In a preferred embodiment, training the waste identification neural network model includes: Obtain a waste sample training dataset; wherein, the training dataset includes several substation scene sample images containing various types of waste, and each substation scene sample image is labeled with the real area coordinates and real category label of the waste; The waste sample training dataset is divided into several batches of training samples according to a preset batch size; Each batch of training samples is input into the waste identification neural network model to be trained for iterative training until the preset number of training rounds is met; wherein, in each iteration, the waste identification neural network model to be trained outputs the predicted waste area coordinates and the predicted category confidence based on the current batch of training samples. Based on the predicted waste area coordinates and the corresponding real area coordinates, a location loss value is calculated using a location regression loss function; based on the predicted category confidence and the corresponding real category label, a classification loss value is calculated using a classification loss function; the location loss value and the classification loss value are weighted and summed to obtain a comprehensive loss value, and a preset optimizer is used to update the network parameters of the waste identification neural network model based on the comprehensive loss value.

[0037] Specifically, to achieve intelligent perception of unstructured environmental anomalies such as waste discharge, this method introduces a waste risk assessment step based on computer vision. This step automatically identifies abnormal waste discharge situations, such as oil stains left by sulfur hexafluoride (SF6) leaks, transformer oil seepage, and illegally dumped sewage, through deep image feature analysis of the monitoring video data, and quantifies these anomalies to generate a waste risk value (denoted as ). Specifically, this can be achieved using traditional rule-based image processing techniques or artificial intelligence techniques based on deep learning.

[0038] In a preferred embodiment, waste identification is performed using traditional image processing techniques based on background subtraction and feature matching. The process first extracts the frame image to be detected from the real-time transmitted monitoring video data at a preset sampling frequency (e.g., 5 frames per second). Considering the variability of outdoor lighting conditions, the system does not use a single fixed background. Instead, for each frame image to be detected, based on the current timestamp or average image brightness information, it matches and obtains a background reference image from a preset background library that most closely matches the lighting environment characteristics of the current frame image to be detected.

[0039] The pre-set background library is an image database built by the system during the initialization phase or through a periodic self-learning update mechanism. It covers images of the normal environmental conditions (i.e., clean backgrounds confirmed to be free of waste emissions) of the substation monitoring area under different time periods (such as early morning, noon, dusk, and night) and different weather conditions (such as sunny, cloudy, and rainy). To improve retrieval efficiency, each background image in the library has its corresponding illumination feature parameters pre-extracted and stored as index tags.

[0040] The specific logic for determining "most similar lighting environment features" is as follows: First, calculate the global lighting feature value of the current frame image to be detected. This feature value can be the average grayscale value, the average value of the luminance component (Y channel), or the value histogram in the HSV color space. Then, compare the calculated feature value with the corresponding feature parameters of each background image in the preset background library, and calculate the numerical difference between the two (e.g., Euclidean distance or absolute difference). Finally, select the background image with the smallest numerical difference or the highest histogram correlation as the background reference image that is most similar to the current lighting environment features. Through this dynamic matching mechanism, the overall lighting difference caused by changes in the sun angle, cloud cover, or artificial lighting can be eliminated to the greatest extent, thereby ensuring that the foreground difference region extracted by subsequent difference operations is indeed from newly added objects (waste) in the environment, rather than false contours caused by changes in light and shadow.

[0041] Subsequently, pixel-level difference operations are performed between the current frame image to be detected and the background reference image. After binarization and morphological filtering, the foreground difference region is extracted. Next, the color features (such as the color mean in HSV space) and texture features (such as the texture complexity of the gray-level co-occurrence matrix) of the foreground difference region are calculated. Since waste such as transformer oil and sewage usually have specific color ranges and relatively smooth surfaces (low texture complexity), if the color mean of the foreground difference region falls within a preset waste color range and the texture complexity is less than a preset liquid texture threshold, it is preliminarily determined that there is a suspected waste region in the current frame image to be detected. To eliminate instantaneous interference such as swaying leaves and flying birds, the system counts the number of consecutive frames to be detected that are determined to have suspected waste regions. If the number of consecutive frames to be detected that have suspected waste regions reaches a preset alarm threshold (e.g., 30 consecutive frames), it is confirmed that an abnormal waste discharge has occurred, and the frame image to be detected that exactly reaches the alarm threshold is selected as the target image. Finally, the pixel area ratio of the suspected waste region in the target image is calculated, and the area ratio is directly determined as the waste risk value R3; if the above consecutive frame number condition is not met or feature matching fails, the waste risk value R3 is assigned to zero.

[0042] In another preferred embodiment, a deep neural network-based intelligent recognition technology is employed to further enhance the robustness of the recognition. To more clearly illustrate the specific construction and operation mechanism of the waste recognition neural network model, this embodiment uses the YOLO (You Only Look Once) series network architecture (e.g., the YOLOv5 model) or the Faster R-CNN network architecture as examples for detailed explanation. Taking the YOLOv5 model as an example, the network hierarchy of the waste recognition neural network model mainly consists of four parts: Input, Backbone, Neck, and Head. The specific operation process is as follows: First, the extracted image to be detected is adaptively scaled and normalized at the input end to adjust it to a uniform size required by the network; then, the processed image data enters the backbone network (e.g., using a CSPDarknet structure), which contains several convolutional layers, batch normalization layers, and activation function layers. Through multiple convolutional and pooling operations, shallow geometric features and deep semantic features of different scales in the image are extracted layer by layer. This process realizes the convolutional feature extraction of the image to be detected; then, the extracted multi-scale feature maps... The feature maps are fed into the neck network (e.g., using an FPN+PAN structure) and fused from different levels through top-down upsampling and bottom-up downsampling paths. This enhances the model's ability to perceive waste targets of different sizes (such as large areas of sewage or small oil stains). Finally, the fused feature maps are input to the detection head, which contains multiple detection layers of different scales. Through convolution operations, the feature maps are mapped to prediction results, and tensor data containing bounding box coordinate parameters, target confidence scores, and class probabilities are directly output, thus completing target localization and recognition confidence generation.

[0043] Based on the aforementioned network structure, the identification process first extracts serialized test frame images from the monitoring video data according to a preset sampling frequency. Then, each test frame image is sequentially input into the preset waste identification neural network model. The model outputs the identification confidence score for the presence of waste in each test frame image and the corresponding coordinates of the waste target area. To ensure the accuracy of the alarm, the system uses a counter to count the number of test frame images whose identification confidence score is continuously greater than a preset effective threshold (e.g., 0.8). The specific counting logic is as follows: when the identification confidence score of the current test frame image is greater than the effective threshold, the counter is incremented; conversely, if the identification confidence score of the current frame image is less than or equal to the effective threshold, the counter is reset to zero. When the counter value accumulates to a preset alarm threshold, an abnormal waste discharge is identified, and the corresponding test frame image that exactly causes the counter value to reach the alarm threshold is immediately selected as the target image. Subsequently, the pixel area ratio of the waste target area in the target image is calculated based on the bounding box coordinates output by the model, and this area ratio is determined as the waste risk value. If the counter never reaches the alarm threshold, then the waste risk value will be... The value is assigned to zero.

[0044] To obtain the aforementioned high-performance waste identification neural network model, this embodiment also discloses a training method for the model. First, sample images of substation scenes containing various types of waste are acquired as a waste sample training dataset. Each sample image is manually or semi-automatically labeled, clearly indicating the true bounding box coordinates and true category labels (e.g., oil, sewage). Next, the training dataset is divided into several batches of training samples according to a preset batch size. Each batch of training samples is input into the waste identification neural network model to be trained for forward propagation and multiple rounds of iterative training are performed until a preset number of training rounds or convergence conditions are met. In each iteration, the model outputs the predicted waste region coordinates and predicted category confidence based on the current batch of training samples. The core of the training process lies in updating parameters using a multi-task loss function: Based on the predicted waste area coordinates and the corresponding ground truth area coordinates, a location regression loss function (e.g., CIoULoss or Smooth L1 Loss) is used to calculate the localization loss value; based on the predicted class confidence and the corresponding ground truth class label, a classification loss function (e.g., Cross Entropy Loss) is used to calculate the classification loss value. Finally, the localization loss value and the classification loss value are weighted and summed to obtain a comprehensive loss value. A preset optimizer (e.g., Adam or SGD) is then used to update the network parameters of the waste identification neural network model based on this comprehensive loss value through backpropagation, until the model achieves high-precision waste identification capabilities.

[0045] Step S5: Select the maximum value among the noise risk value, the electromagnetic radiation risk value, and the waste risk value, and determine the maximum value as the comprehensive risk value; compare the comprehensive risk value with the preset risk classification threshold to determine the substation environmental risk level; generate corresponding alarm control signals and operation and maintenance strategies based on the substation environmental risk level.

[0046] In a preferred embodiment, the risk classification threshold includes a first risk threshold, a second risk threshold, and a third risk threshold; the first risk threshold is less than the second risk threshold, and the second risk threshold is less than the third risk threshold; The step of comparing the comprehensive risk value with a preset risk classification threshold to determine the environmental risk level of the substation includes: If the comprehensive risk value is less than the first risk threshold, the environmental risk level of the substation is determined to be acceptable. If the comprehensive risk value is greater than or equal to the first risk threshold and less than the second risk threshold, then the environmental risk level of the substation is determined to be a minor risk level. If the comprehensive risk value is greater than or equal to the second risk threshold and less than the third risk threshold, then the environmental risk level of the substation is determined to be a medium risk level. If the comprehensive risk value is greater than or equal to the third risk threshold, the environmental risk level of the substation is determined to be a severe risk level.

[0047] Specifically, in the implementation process of the substation environmental risk monitoring and early warning method, after calculating the noise risk value, the electromagnetic radiation risk value, and the waste risk value respectively, a comprehensive risk assessment step is performed. To ensure the rigor of environmental safety management, this method adopts the "maximum value principle" to integrate multi-dimensional risk indicators, that is, selecting the maximum value among the noise risk value, the electromagnetic radiation risk value, and the waste risk value, and directly determining this maximum value as the comprehensive risk value (denoted as ). The comprehensive risk value The computational logic can be expressed as follows: This logic ensures that if any single environmental element significantly exceeds the standard, the overall environmental assessment will reflect a high-risk state, avoiding the risk masking problem that may be caused by the weighted average algorithm. Subsequently, the calculated comprehensive risk value... The risk level of the substation is determined by comparing it with the risk classification thresholds preset in the database.

[0048] In a preferred embodiment, the risk grading threshold comprises three incremental numerical standards, namely a first risk threshold (denoted as...). ), and the second risk threshold (denoted as ) and the third risk threshold (denoted as And all three conditions are met. The numerical relationship. The specific logic for determining the environmental risk level of a substation is as follows: if the comprehensive risk value Less than the first risk threshold If the environmental risk level of the substation is determined to be "acceptable," it indicates that the current environmental condition is good and no intervention is required; if the comprehensive risk value is... Falling at the first risk threshold (inclusive) and the second risk threshold If the risk level is between (excluding) and (not including), it is classified as "minor risk level"; if the comprehensive risk value is... Falling at the second risk threshold (inclusive) and the third risk threshold If the risk level is between (excluding) and (not including), it is classified as "medium risk level"; if the comprehensive risk value is... If the risk level is greater than or equal to the third risk threshold F3, it is determined to be "severe risk level".

[0049] Based on the determined environmental risk level of the substation, this method further generates graded alarm control signals and targeted operation and maintenance strategies. For the alarm control signals, the system drives the on-site audible and visual alarm devices to perform differentiated actions according to the risk level: when the risk level is determined to be minor, a yellow alarm control signal is generated to illuminate a yellow warning light; when the risk level is determined to be medium, an orange alarm control signal is generated to illuminate an orange warning light; and when the risk level is determined to be severe, a red alarm control signal is generated to illuminate a red warning light. Regarding the generation of the operation and maintenance strategies, this method backtracks the comprehensive risk value. Accurate matching of data source categories: if the comprehensive risk value If the noise risk value is derived from the noise risk value, then a noise risk control strategy (such as checking sound barriers) is retrieved from the strategy library; if the comprehensive risk value... If the electromagnetic radiation risk value is derived from the electromagnetic radiation risk value, then the electromagnetic risk prevention and control strategy will be invoked; if the comprehensive risk value is... If the risk value originates from the waste, a waste risk control strategy is invoked (such as initiating wastewater treatment or dispatching cleaning personnel). Through the above steps, a closed-loop management system covering the entire process from risk identification and quantification to decision support is achieved.

[0050] Based on the above method embodiments, the present invention provides corresponding apparatus embodiments.

[0051] like Figure 2 As shown, an embodiment of the present invention provides a substation environmental risk monitoring and early warning device based on multi-dimensional data fusion, including: a data acquisition module, a noise risk assessment module, an electromagnetic radiation risk assessment module, a waste risk assessment module, and a comprehensive early warning and control module; The data acquisition module is used to acquire noise data, environmental meteorological parameter data, electromagnetic radiation data, and monitoring video data of the substation. The noise risk assessment module is used to calculate the predicted noise value at the target residential area based on the noise data, the environmental meteorological parameter data, and the preset propagation distance parameter, using an acoustic propagation attenuation model; compare the predicted noise value with the preset noise emission limit to generate a noise risk value; wherein, the propagation distance parameter represents the distance between the substation and the target residential area; The electromagnetic radiation risk assessment module is used to calculate the predicted electromagnetic radiation value at the target residential area based on the electromagnetic radiation data, the propagation distance parameter, and the preset monitoring end distance parameter; compare the predicted electromagnetic radiation value with the preset electromagnetic control limit to generate an electromagnetic radiation risk value; wherein, the monitoring end distance parameter represents the distance between the electromagnetic radiation data acquisition location and the radiation source in the substation. The waste risk assessment module is used to perform image feature analysis on the monitoring video data, identify abnormal waste discharge situations, and generate waste risk values. The integrated early warning and control module is used to select the maximum value among the noise risk value, the electromagnetic radiation risk value, and the waste risk value, and determine the maximum value as the integrated risk value; compare the integrated risk value with a preset risk classification threshold to determine the substation environmental risk level; and generate corresponding alarm control signals and operation and maintenance strategies based on the substation environmental risk level.

[0052] It should be noted that the embodiments of the device described above correspond to the embodiments of the present invention described above, and can realize the substation environmental risk monitoring and early warning method based on multi-dimensional data fusion described in any one of the above embodiments of the present invention. Furthermore, the embodiments of the device described above are merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between modules indicates that they have a communication connection, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without creative effort.

[0053] Based on the above-described method embodiments of the present invention, a corresponding embodiment of an electronic device is provided.

[0054] An embodiment of the present invention provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the substation environmental risk monitoring and early warning method based on multi-dimensional data fusion as described in any one of the present invention, or the processor executes the computer program to implement the functions of each module in the above-described device embodiments.

[0055] For example, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the terminal device.

[0056] The terminal device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0057] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting all parts of the terminal device via various interfaces and lines.

[0058] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the terminal device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, applications required for at least one function, etc.; the data storage area may store data created based on the use of the mobile phone, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital card (SD card), flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0059] Based on the above method embodiments, the present invention provides corresponding storage medium embodiments; Another embodiment of the present invention provides a storage medium including a stored computer program, wherein, when the computer program is running, it controls the device where the storage medium is located to execute any of the above-described substation environmental risk monitoring and early warning methods based on multi-dimensional data fusion of the present invention.

[0060] The aforementioned storage medium is a computer-readable storage medium, and the computer program includes computer program code, which may be in the form of source code, object code, executable file, or certain intermediate forms. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0061] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of those different embodiments or examples.

[0062] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for monitoring and early warning of environmental risks in substations based on multi-dimensional data fusion, characterized in that, include: Acquire noise data, environmental meteorological parameter data, electromagnetic radiation data, and monitoring video data of the substation; Based on the noise data, the environmental meteorological parameter data, and the preset propagation distance parameter, the predicted noise value at the target residential area is calculated using an acoustic propagation attenuation model; the predicted noise value is compared with the preset noise emission limit to generate a noise risk value; wherein, the propagation distance parameter represents the distance between the substation and the target residential area; Based on the electromagnetic radiation data, the propagation distance parameter, and the preset monitoring terminal distance parameter, the predicted electromagnetic radiation value at the target residential area is calculated; the predicted electromagnetic radiation value is compared with the preset electromagnetic control limit to generate an electromagnetic radiation risk value; wherein, the monitoring terminal distance parameter represents the distance between the electromagnetic radiation data acquisition location and the radiation source in the substation. Image feature analysis is performed on the monitoring video data to identify abnormal waste discharge situations and generate waste risk values; The maximum value among the noise risk value, the electromagnetic radiation risk value, and the waste risk value is selected and determined as the comprehensive risk value; the comprehensive risk value is compared with a preset risk classification threshold to determine the substation environmental risk level; and corresponding alarm control signals and operation and maintenance strategies are generated based on the substation environmental risk level.

2. The substation environmental risk monitoring and early warning method based on multi-dimensional data fusion as described in claim 1, characterized in that, The predicted noise value at the target residential area is calculated using an acoustic propagation attenuation model based on the noise data, the environmental meteorological parameter data, and the preset propagation distance parameter. The predicted noise value is compared with a preset noise emission limit to generate a noise risk value, including: Based on the propagation distance parameter, the geometric divergence attenuation of the sound wave during propagation is calculated; The atmospheric absorption attenuation coefficient is determined based on the environmental meteorological parameter data, and the atmospheric absorption attenuation of the sound wave during propagation is calculated using the propagation distance parameter and the atmospheric absorption attenuation coefficient. The noise data is attenuated and corrected using the geometric divergence attenuation and the atmospheric absorption attenuation to obtain the predicted noise value at the target residential area. If the predicted noise value is greater than the preset noise emission limit, then the first difference between the predicted noise value and the noise emission limit is calculated, and the ratio of the first difference to the noise emission limit is determined as the noise risk value. If the predicted noise value is less than or equal to the noise emission limit, then the noise risk value is set to zero.

3. The substation environmental risk monitoring and early warning method based on multi-dimensional data fusion as described in claim 1, characterized in that, The predicted electromagnetic radiation value at the target residential area is calculated based on the electromagnetic radiation data, the propagation distance parameter, and the preset monitoring end distance parameter. The predicted electromagnetic radiation value is compared with a preset electromagnetic control limit to generate an electromagnetic radiation risk value, including: Based on the preset monitoring end distance parameter and the propagation distance parameter, an electromagnetic attenuation coefficient characterizing the distance attenuation characteristics is constructed; The electromagnetic radiation data is attenuated and corrected using the electromagnetic attenuation coefficient to obtain the predicted electromagnetic radiation value at the target residential area. If the predicted electromagnetic radiation value is greater than the preset electromagnetic control limit, then the second difference between the predicted electromagnetic radiation value and the electromagnetic control limit is calculated, and the ratio of the second difference to the electromagnetic control limit is determined as the electromagnetic radiation risk value. If the predicted electromagnetic radiation value is less than or equal to the electromagnetic control limit, then the electromagnetic radiation risk value is set to zero.

4. The substation environmental risk monitoring and early warning method based on multi-dimensional data fusion as described in claim 1, characterized in that, The step of performing image feature analysis on the monitored video data to identify abnormal waste discharge situations and generate waste risk values ​​includes: The frame image to be detected is extracted from the monitoring video data according to a preset sampling frequency; For each frame image to be detected, a background reference image that matches the lighting environment features of the current frame image to be detected is obtained, and the current frame image to be detected and the background reference image are subjected to a difference operation to extract the foreground difference region; Calculate the color mean and texture complexity of the foreground difference region; If the average color value of the foreground difference region falls within the preset waste color range, and the texture complexity of the foreground difference region is less than the preset liquid texture threshold, then it is determined that there is a suspected waste region in the current frame image to be detected. The system counts the number of consecutive frames of images identified as having suspected waste areas. If the number of consecutive frames of images with suspected waste areas reaches a preset alarm threshold, the frame of images that exactly reaches the alarm threshold is selected as the target image. The area ratio of the suspected waste area in the target image is calculated, and the area ratio is determined as the waste risk value. Otherwise, assign a waste risk value of zero.

5. The substation environmental risk monitoring and early warning method based on multi-dimensional data fusion as described in claim 1, characterized in that, The step of performing image feature analysis on the monitored video data to identify abnormal waste discharge situations and generate waste risk values ​​includes: The serialized frame images to be detected are extracted from the monitoring video data according to a preset sampling frequency; Each of the images to be detected is sequentially input into a preset waste identification neural network model, so that the waste identification neural network model generates the recognition confidence and waste target area of ​​each image to be detected based on the images to be detected. A counter is used to count the number of frames of the image to be detected whose recognition confidence is continuously greater than a preset effective threshold; wherein, when the recognition confidence of the current frame image is greater than the effective threshold, the counter is incremented by one; when the recognition confidence of the current frame image is less than or equal to the effective threshold, the counter is reset to zero. When the counter value accumulates to a preset alarm threshold, an abnormal waste discharge is identified, and the corresponding frame image to be detected that just causes the counter value to reach the alarm threshold is selected as the target image. Calculate the area ratio of the waste target region in the target image, and determine the area ratio as the waste risk value; If the counter does not reach the alarm threshold, the waste risk value will be set to zero.

6. The substation environmental risk monitoring and early warning method based on multi-dimensional data fusion as described in claim 5, characterized in that, The training of the waste identification neural network model includes: Obtain a waste sample training dataset; wherein, the training dataset includes several substation scene sample images containing various types of waste, and each substation scene sample image is labeled with the real area coordinates and real category label of the waste; The waste sample training dataset is divided into several batches of training samples according to a preset batch size; Each batch of training samples is input into the waste identification neural network model to be trained for iterative training until the preset number of training rounds is met; wherein, in each iteration, the waste identification neural network model to be trained outputs the predicted waste area coordinates and the predicted category confidence based on the current batch of training samples. Based on the predicted waste area coordinates and the corresponding real area coordinates, a location loss value is calculated using a location regression loss function; based on the predicted category confidence and the corresponding real category label, a classification loss value is calculated using a classification loss function; the location loss value and the classification loss value are weighted and summed to obtain a comprehensive loss value, and a preset optimizer is used to update the network parameters of the waste identification neural network model based on the comprehensive loss value.

7. The substation environmental risk monitoring and early warning method based on multi-dimensional data fusion as described in claim 1, characterized in that, The risk classification thresholds include a first risk threshold, a second risk threshold, and a third risk threshold. The first risk threshold is less than the second risk threshold, and the second risk threshold is less than the third risk threshold; The step of comparing the comprehensive risk value with a preset risk classification threshold to determine the environmental risk level of the substation includes: If the comprehensive risk value is less than the first risk threshold, the environmental risk level of the substation is determined to be acceptable. If the comprehensive risk value is greater than or equal to the first risk threshold and less than the second risk threshold, then the environmental risk level of the substation is determined to be a minor risk level. If the comprehensive risk value is greater than or equal to the second risk threshold and less than the third risk threshold, then the environmental risk level of the substation is determined to be a medium risk level. If the comprehensive risk value is greater than or equal to the third risk threshold, the environmental risk level of the substation is determined to be a severe risk level.

8. A substation environmental risk monitoring and early warning device based on multi-dimensional data fusion, characterized in that, include: The system includes a data acquisition module, a noise risk assessment module, an electromagnetic radiation risk assessment module, a waste risk assessment module, and a comprehensive early warning and control module. The data acquisition module is used to acquire noise data, environmental meteorological parameter data, electromagnetic radiation data, and monitoring video data of the substation. The noise risk assessment module is used to calculate the predicted noise value at the target residential area based on the noise data, the environmental meteorological parameter data, and the preset propagation distance parameter, using an acoustic propagation attenuation model; compare the predicted noise value with the preset noise emission limit to generate a noise risk value; wherein, the propagation distance parameter represents the distance between the substation and the target residential area; The electromagnetic radiation risk assessment module is used to calculate the predicted electromagnetic radiation value at the target residential area based on the electromagnetic radiation data, the propagation distance parameter, and the preset monitoring end distance parameter; compare the predicted electromagnetic radiation value with the preset electromagnetic control limit to generate an electromagnetic radiation risk value; wherein, the monitoring end distance parameter represents the distance between the electromagnetic radiation data acquisition location and the radiation source in the substation. The waste risk assessment module is used to perform image feature analysis on the monitoring video data, identify abnormal waste discharge situations, and generate waste risk values. The integrated early warning and control module is used to select the maximum value among the noise risk value, the electromagnetic radiation risk value, and the waste risk value, and determine the maximum value as the integrated risk value; compare the integrated risk value with a preset risk classification threshold to determine the substation environmental risk level; and generate corresponding alarm control signals and operation and maintenance strategies based on the substation environmental risk level.

9. An electronic device, characterized in that, The method includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the substation environmental risk monitoring and early warning method based on multidimensional data fusion as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium includes a stored computer program, wherein, when the computer program is running, it controls the device where the storage medium is located to execute the substation environmental risk monitoring and early warning method based on multi-dimensional data fusion as described in any one of claims 1 to 7.

Images